[clang] [llvm] [AMDGPU][GFX12.5] Reimplement monitor load as an atomic operation (PR #177343)

[Joseph Huber via cfe-commits]

================
@@ -341,6 +348,29 @@ bool SemaAMDGPU::CheckAMDGCNBuiltinFunctionCall(unsigned BuiltinID,
   return false;
 }
 
+bool SemaAMDGPU::checkAtomicOrderingCABIArg(CallExpr *TheCall, unsigned ArgIdx,
+                                            bool MayLoad, bool MayStore) {
+  Expr *AtomicOrdArg = TheCall->getArg(ArgIdx);
+  Expr::EvalResult AtomicOrdArgRes;
+  if (!AtomicOrdArg->EvaluateAsInt(AtomicOrdArgRes, getASTContext()))
----------------
jhuber6 wrote:

Yeah you need to skip out of Sema if it's type dependent usually. It's why it's much nicer if don't need custom typechecking, but I don't know if `_Constant const char*` actually works.

https://github.com/llvm/llvm-project/pull/177343